Some Costs of
Reliability Limits
Douglas R. Hale and Thomas Leckey
DHALE@eia.doe.gov
tleckey@eia.doe.gov
U.S. Energy Information Administration
Washington, DC 20585 USA
Thomas J. Overbye
Overbye@ece.uiuc.edu
University of Illinois at Urbana-Champaign
Urbana, IL 61801 USA
Abstract
To ensure that the electrical grid will continue to operate even when generators and lines fail,
electricity system managers restrict power flows between areas to levels below the physical capability
of the lines. Electricity traders have complained that these safety margins, which are formalized in
area interchange limits, can preclude profitable trades. Line limits can also protect local markets from
outside competition. This paper quantifies some of the costs of area interchange limits constraining
power flows between areas in the Eastern Interconnect in the summer of 2000. Recognizing the
hidden costs and competitive effects of line limits is a prerequisite to balancing reliability goals
against their cost.
1. Introduction
Utilities, Independent System Operators (ISOs) and National Electric Reliability Council (NERC)
reliability areas rely on excess generator capacity and redundant line capability to ensure electricity
demand, including losses, does not exceed generation. If it does, voltages and frequency may drop,
possibly causing blackouts. System managers have traditionally required generation capacity that
exceeds expected peak demand. The “safety margin” in the Northeast Power Coordinating Council
(NPCC)1 is currently about 15%; suppliers (load serving entities) in the PJM Interconnect2 are
required to have a 17% excess of generation over their peak demand.
Similarly, system operators, usually control areas, enforce power flow limits on lines with their
neighbors that include safety margins derived from contingency analyses. In a contingency analysis,
engineers present a model of the electrical system with hypothetical demand conditions and a base
1
NPCC consists of New York State, the New England States, and the Ontario, Quebec and Maritime Provinces.
PJM is an acronym for Pennsylvania, (New) Jersey, Maryland, and the system also includes Washington, DC and the
Delmarva Peninsula.
2
case of operating generators and lines. Large generators and major lines are then taken off line one at a
time to mimic unplanned outages. This is called an n-1 contingency analysis: all but 1 of the n pieces
of major equipment in the electrical system is assumed to operate normally. The analysts note those
operating regimes that cause failure of other large lines, potentially resulting in cascading blackouts.
Through a planning procedure, they preclude catastrophic failures, essentially “outlawing” failed
operating regimes, by de-rating vulnerable power lines. The line limits that are imposed to ensure that
the system continues to operate after a failure are called area interchange limits, n-1 limits,
contingency limits, reliability limits or some similar term. System operators treat these area
interchange limits as if they were actual physical limits when they dispatch the system, thus reducing
the chance of outage induced blackouts. 3
The n-1 approach to achieving reliable electricity supply has been effective: Large area blackouts have
been rare in the United States since the North American Reliability Council was established in 1968.
But in recent years, some control areas have canceled large volumes of otherwise profitable trade to
protect the lines against hypothetical events.4 Unfortunately since the probability of blackouts is not
calculated as part of the n-1 approach,5 it is not possible to know whether canceling these trades
actually reduced the risk of blackout. Another problem is once limits are set they are not dynamically
updated to reflect the real time evolution of demand, supply, equipment failures and weather.
Consequently trades could be disallowed when there is no realistic chance of blackout and allowed
when there is. The emergence of competition has also led to new complaints that reliability limits are
being manipulated to protect incumbent suppliers from outside competition.
This paper quantifies the short term operating costs of the area interchange limits reported by NERC
for the Eastern Interconnect in the summer of 2000. The Eastern Interconnect is the US east of the
Rocky Mountains, exclusive of Texas, and several Canadian Providences. The paper also illustrates
how more realistic line limits can act to protect local markets from high prices.
The paper is organized as follows. Section 2 presents the model used to make the cost and price
calculations. Section 3 shows that managing the Eastern Interconnect at its physical potential would
bring significant benefits. Under peak summer demand conditions across the Interconnect, system
operating costs could be reduced almost 4 percent by loosening area interchange (reliability) limits.
This is a large gain considering that to meet peak demand simultaneously requires almost all of the
available generator capacity. It is on the order of the savings, also about 4 percent, estimated for the
restructuring of the United Kingdom’s electricity industry.6 When demand is less and more relatively
cheap capacity is available the potential cost reductions would be larger. At lower demand levels, the
savings increase to 6.6%.
3
Note that if a major piece of equipment has already failed, the n-2 limits become the relevant constraints.
The Federal Energy Regulatory Commission has endorsed the use of transmission loading relief orders to individual
generators to keep line flows below area interchange limits. These orders are based on a transaction's "priority" and not on
its economic value. In addition, those generators who have a priority cannot sell it to others who are willing to pay. A
summary of recent curtailments appears in Transmission Constraint Study, Presentation of FERC Staff to the Commission,
December 19, 2001.
5 The probability of system failure depends not only on what equipment fails but also on what information is available to
operators and what they do with it. These considerations are currently resolved by operator judgment.
6 D.M. Newbery & M.G. Pollitt "The Restructuring and Privatisation of the CEGB: Was It Worth It?" Journal of
Industrial Economics, 1997, 45(3): 269-303.
4
2
Section 4 illustrates how the choice of area interchange limit can dominate pricing. As shown by the
examples of New England and Florida, transmission managers such as Regional Transmission
Organizations (RTOs) will have the power to eliminate the gains from trade and competition, which
we estimate as 24% for New England and 5% for Florida, by their choice of contingency (trade)
limits. Section 5 discusses congestion problems around a load pocket, New York City, and Section 6
is the conclusion.
The Eastern Interconnect Model
The short term operating cost of area interchange (reliability) limits is the difference between the least
cost of meeting demand while enforcing the limits and the least cost of meeting demand consistent
with the transmission system’s physical capabilities. Specifically we ask, if the Eastern Interconnect
were managed to meet demand at least cost while respecting all electrical constraints, but not
necessarily all of the area interchange limits, would there be significant short-term benefits? To
estimate the costs of interchange limits we assume that the existing utility-based systems minimize
costs within their borders but restrict trade with their neighbors to honor the limits reported by NERC.
The current situation is called the “restricted trade” regime. By comparison, the least cost alternative
allows any economic (lower cost) trade that is physically feasible. This is called “unrestricted trade.”
Since regional markets are not perfectly competitive, their actual costs are understated by this
estimate. Consequently the actual costs of area interchange limits are higher than reported here.
Trade reduces the cost of meeting electricity demand by replacing generation from high cost, but local,
generators with generation from lower cost, distant suppliers. How much cost is reduced and how
much electricity can be transmitted from low cost to higher cost areas depends upon the relative costs
of hundreds of generators and the details of the transmission system (lines, capacitors, transformers
and so forth) connecting generators to electricity users. The opportunities for trade are also limited by
constraints placed on line flows in order to promote reliability.
Computing the cost of meeting demand under “restricted” and “unrestricted” trade in the face of this
complexity requires a mathematical model. The model used in this study consists of four components:
a description of the electrical properties of essentially all of the generators and higher voltage power
lines in the Eastern Interconnect; estimated operating costs of most of the generator capacity; the
contingency limits; and, an algorithm for calculating the least cost of meeting demand consistent with
the physics of the transmission system. The Eastern Interconnect includes the National Electrical
Reliability Council’s (NERC’s) regions shown in Figure 1 except for WSSC and ERCOT. It is roughly
the United States east of the Rocky Mountains and excluding Texas. WSSC and ERCOT are
electrically separate from the East and so very little electricity crosses their boundaries.
The electrical model for the study is based on the NERC’s summer 2000 Base Case model of the
Eastern Interconnect. NERC’s Multiregional Modeling Working Group built the Summer 2000 Base
Case from regional power flow models provided by NERC’s members. This model is intended to
“…realistically simulate bulk electric system behavior.” The Committee’s guidance for component
models ensures that the integrated model is realistic:
3
Of paramount importance in this effort is the detail in which the various systems are modeled.
The detail included in each system model must be adequate for all inter and intra-regional
study activities but not necessarily as detailed as required for internal studies. This means that
each system model should include sufficient detail to ensure that power transfers or
contingencies can be realistically simulated.7
The summer 2000 Base Case consists of 33,528 buses, 22,812 loads, 5,312 generators, 2,361 switched
shunts, 45,421 AC lines/transformers (lines), 10 DC lines, and 107 control areas. Total load at peak is
536.4 GW. A significant portion of the modeled case lies outside the United States portion of the
Eastern Interconnect, which is the primary focus of this analysis. Modeled peak load in the United
States is about 510 GW, around 30 GW less than modeled in the entire case. Similarly, about 4,740
generators and 31,000 of the buses are located in the United States. Of the 107 areas represented, 11
are in Canada, and another 7 represent equivalenced power flows to areas in the Western Systems
Coordinating Council (WSCC) and the Electric Reliability Council of Texas (ERCOT). These 89
areas of interest comprise 9
NERC regions in the Eastern
Figure 1
Interconnect (Figure 1). The
NERC flowgates (line limits)
described on the NERC
Market Re-dispatch website
were also included in the
electrical model.8
Short run operating costs for
generators are essential in
analyzing optimal power
flows.
Unfortunately, the
NERC 2000 case does not
report operating costs for
generators,
nor
is
it
straightforward to match the
generators listed
in the
NERC 2000 case with those
reported on EIA’s inventory
survey forms.9
Table 1
reports the NERC regional
distribution of generators modeled and identified.
Overall, 86 percent of the generators were
identified which enabled the technological character to be determined for 96 percent of the capacity.
7
NERC, Multiregional Modeling Working Group Procedural Manual, June 2000, page 2
The original NERC case contained a large number of electrical violations, including overloaded lines and substandard
voltage in some areas. Typically the violations occurred in remote locations and involved small amounts of energy. Most
of these violations were fixed by correcting apparently inaccurate data. Details are available from Thomas Overbye.
9 Utility units are detailed on the EIA-860A, “Annual Electric Generator Report – Utility” and non-utility units are
reported on the EIA-860B, “Annual Electric Generator Report – Nonutility.”
8
4
Identification
was
most difficult in New
England and New
York where small
hydroelectric,
cogeneration,
and
non-utility generators
are most prevalent.
Table 1. Generating Units with Modeled Costs, Number and Capacity by NERC Region
Units
Capacity (GW)
Region
Modeled
Identified
Percent Modeled Identified Percent
ECAR
551
466
84.6%
107.1
102.4
95.6%
FRCC
309
289
93.5%
38.3
37.6
98.2%
MAIN
435
361
83.0%
56.1
53.1
94.5%
MAPP
512
441
86.1%
36.9
35.5
96.1%
NEPOOL
475
334
70.3%
26.0
23.7
90.9%
NYPP
387
286
73.9%
34.4
32.0
93.1%
PJM
484
422
87.2%
59.0
57.0
96.6%
SERC
1209
1117
92.4%
170.8
164.3
96.2%
SPP
362
329
90.9%
45.6
44.1
96.9%
East Interconnect
4724
4045
85.6%
574.2
549.7
95.7%
Our approach to
estimating operating
costs per kilowatthour for fossil units was to use the cost of fuel as reported on the FERC Form 42310
for the month of June 2000, and then to adjust the fuel cost by multiplying by the heat rate.11
Representative heat rates were developed from a variety of sources, most often the EIA-76712 for large
steam turbines, and the FERC Form 423 in combination with the EIA-759.13 These costs were
developed at the plant level, and then assigned to the appropriate units. About 69% of the capacity
was assigned costs using this method. Where it was possible to identify individual steam units
burning coal and another fossil fuel, unit level estimates of costs associated with the various fuels were
developed. Four percent of the capacity was assigned these unit level costs.14 Because many of the
generating units modeled either fell below the reporting threshold or were outside the survey frame
(non-utilities), a costing alternative was developed. Fossil generators without reported plant level
costs were assigned state-level averages derived from the FERC Form 423, and adjusted using plantspecific heat rates. Another four percent of capacity was assigned using this alternative methodology.
Costs for nuclear units (15 percent) were developed from the FERC Form I, for 1999, adjusted to 2000
dollars.15 These costs were uniformly low.
Hydroelectric and pumped storage plants, about
Table 2. Capacity and Load by NERC Region
Region
Capacity (GW) Load (GW)
Ratio
eight percent of the capacity, were assigned the
ECAR
107.1
93.0
1.15
same nominal cost, low enough to make them
FRCC
38.3
36.5
1.05
inframarginal.
MAIN
56.1
48.8
1.15
Table 2 reports the capacity modeled16 in the NERC
2000 case, as well as the forecasted demand (load)
by NERC region. The ratio of regional capacity to
regional load is generally indicative of the reserve
capacity, the ability of a region to serve its load, but
it may also indicate the ability of a region to serve
10
MAPP
NEPOOL
NYPP
PJM
SERC
SPP
East. Int.
36.9
26.0
34.4
59.0
170.8
45.6
574.2
35.5
22.9
28.5
49.5
158.0
37.4
510.1
1.04
1.14
1.21
1.19
1.08
1.22
1.13
“Monthly Cost and Quality of Fuels for Electric Plants.” Reporting threshold is 50 MW at the plant.
$/MMBtu times thousand BTU per kwh, yielding either mills/kWh or $/MWh.
12 “Steam-Electric Plant Operation and Design Report.”
13 “Monthly Power Plant Report.”
14 This differentiation was especially important in PJM, which has about a dozen plants where some of the units burn coal,
while associated units burn gas or petroleum.
15 Inflated about two percent.
16 More than twenty GW of this capacity was originally modeled as off-line, or uncommitted, so in a dispatch the ratios
are more stringent than shown in Table 2.
11
5
load elsewhere. By this measure, SERC and ECAR, significant areas of base load capacity appear to
be “tight” power pools, with relatively slim operating reserves above peak load requirements.
Conversely, New England and New York, two regions characterized by higher than average electricity
prices and which presumably would benefit from trade, appear to have adequate reserves. Florida has
little spare capacity in the NERC 2000 case.
Table 3 displays the regional breakdown and technology of those generators for which cost estimates
were developed and included in the model. About 65% of the capacity that could be assigned a cost is
fossil steam, burning coal, natural gas, residual fuel, or petroleum coke. On average, steam units fired
by coal had costs roughly one-fourth of those fired by either natural gas or petroleum.17 Short run
Table 3. Capacity and Number of Units by Technology Type and Region
Region
ECAR
FRCC
MAIN
MAPP
NEPOOL
NYPP
PJM
SERC
SPP
East Interconnect
Combined Cycle Gas Turbine
Units
MW
Units MW
104 5,404
4
1,620
128 5,672
37
4,524
0
0
110 6,173
5
359
107 4,146
42
3,057
53 2,225
60 2,520
54
3,527
157 8,633
41
3,231
279 14,224
58
5,410
63 3,612
11
1,277
252 23,004 1,061 52,609
Fossil Steam
Nuclear
Units
MW
Units
MW
299 86,362
6 5,732
98 23,035
5 3,954
170 32,540
17 13,305
159 23,569
6 3,580
76 10,671
5 4,324
94 15,862
6 4,922
161 28,862
13 13,035
374 91,742
32 31,063
155 34,994
1 1,164
1,586 347,637
91 81,079
Hydroelectric and
Pumped Storage
Other
Total
Units
MW
Units MW Units
MW
38
2,901
15
378
466 102,397
4
47
17
383
289
37,615
61
975
3
58
361
53,051
69
2,770
95 1,065
441
35,489
118
2,986
40
416
334
23,678
70
5,115
2
50
286
31,996
40
2,887
10
351
422
56,999
373
21,762
1
77 1,117 164,278
76
2,488
23
615
329
44,150
849
41,930
206 3,393 4,045 549,652
marginal costs for nuclear units were also low. Most of the combined cycle units burned relatively
costly natural gas in June 2000, but the efficiency of these units worked to reduce the estimated
operating costs overall. Generally, the gas turbines were the most costly units dispatched in the
model.18 Hydroelectric and pumped storage units were dispatched at a uniformly low cost on the
assumption that they would run during the peak hour. Where reasonable estimates of costs could not
be developed, unit output was fixed at the level of the original NERC 2000 filing, and the generator
was not applied in the optimization algorithm.
Because regions have different mixes of generators, each region ends up with a different estimated
cost profile (Table 4). ECAR, SERC, MAIN, and MAPP possess large amounts of coal and nuclear
baseload capacity, making those regions the lowest cost generation resources in the aggregate. NYPP
and PJM, with larger shares of gas and petroleum steam units, show higher cost profiles. Florida, with
very little coal capacity has the highest short run operating costs on average. The cost profile for both
NEPOOL and NYPP includes significant hydroelectric resources, which are made available at
relatively low cost: under more realistic unit commitment, costs in these areas would be higher than
indicated here. In general, as the model optimizes to reduce cost over large areas, we would expect to
see costs fall in the higher cost areas, though localized transmission constraints could prevent that
outcome in some areas.
17
http://www.eia.doe.gov/cneaf/electricity/epav1/ta20p1.html
“Other” units are mostly internal combustion, consuming gas. A few are MSW (relatively expensive) and wind
(relatively cheap).
18
6
In order to calculate the costs of contingency
limits it is necessary to calculate the least cost in
each trading area when those limits are honored.
NERC’s Summer 2000 case reports the relevant
contingency flow limits. Table 5 reports a few
of the specific flow limits across the trading
areas.19 Negative values indicate imports.
Table 4. Weighted Average Cost of Generation by Region
(nominal dollars/MWh)
Region
Capacity (GW)
$/MWh
FRCC
37.6
38.22
NEPOOL
23.7
34.33
SPP
44.1
32.11
PJM
57.0
30.95
NYPP
32.0
29.37
MAIN
53.1
24.48
MAPP
35.5
22.79
SERC
164.3
21.84
ECAR
102.4
16.45
East Interconnect
549.7
24.96
The physical description of the electrical
network and the generator cost data are inputs
to an algorithm that calculates the least cost of
meeting demand. The algorithm determines the
output of each and every generator in such a way that overall cost of meeting demand is as low as
possible consistent with the laws of physics, the capabilities of the transmission system, generator
availability and capacity and, in the restricted trade regime, administrative limits on each region’s
imports. In the free trade regime there are no administrative limits on imports, just physical limits.
Large non-linear minimum cost problems are hard to solve. Our approach was to solve the non-linear
alternating current equations (i.e., the power flow equations) for an initial trial solution, note the line
and voltage violations, and make a linear approximation to the power flow equations. We then solved
for the minimum cost of a linear version of the non-linear electrical model in such a way as to remove
the violations. This resulted in another “solution” for generator outputs. That solution was given to the
electrical equations and the process started again.
These iterations continued until the solution coming
Table 5. Selected Flow Limits, Summer 2000 Peak
from the cost minimization model was consistent with
Net Scheduled
Transactions
MW
Area
all of the electrical equations.
Least cost operation is an attribute of a perfectly
competitive market.20 Schweppe and his colleagues
showed precisely how least cost operation (net
benefit
maximization)
defines
competitive
21
equilibrium in an electricity market. They showed
the competitive price at any location is equal to the
system cost of one more unit of production, plus the
value of the additional losses incurred in transmitting
power to the location and the cost of the increased
line congestion brought about by increased
consumption. In other words, price at any location is
just the system marginal cost of production plus the
19
PJM
Southern
FRCC
TVA
NY ISO
Entergy
ISO NE
AEP
Commonwealth Edison
Consumers (MI)
Duke
VA Power
First Energy
Cinergy
Ameren
124
598
-2268
-177
-1585
-111
-2526
-543
-435
-508
144
352
-1815
239
-711
FERC Form 715, planning contingencies for 2000.
Least cost operation is also consistent with operating philosophy of regulated utilities. The difference is that regulated
utilities price at average cost and the competitive price is marginal cost. When demand is responsive to price, the levels
and patterns of demand will change. Also, competitors will invest differently from regulated utilities whose returns on
investment are protected by its regulators.
21 Schweppe, Fred, Caramanis, Tabors and Bohn, Spot Pricing of Electricity, Klur Academic Publishers, 1988
20
7
value of induced losses plus the cost of induced congestion.
When congestion and losses are significant, some low cost generators cannot be fully used. Delivering
their full output levels to the grid would cause serious physical damage, perhaps blackouts. In those
circumstances, operators replace generation from cheaper facilities with higher cost generation. This
causes the system cost of supplying an additional unit of electricity to increase, contributing to a
system wide increase in electricity’s price.
Prices typically vary by location because the losses and congestion induced by transmission depend on
where electricity is being demanded. These location dependent prices are called Locational Marginal
Prices (LMPs). Losses and congestion themselves depend on how others are using the transmission
system. For example, when demand is low, lines will usually have excess capacity, and congestion
costs would likely be negligible or absent entirely. If an increase in one customer’s demand is enough
to congest a line, all customers may have to bear congestion charges.22 These external effects of
individuals’ production and consumption decisions are included in the results reported below. But, in
the real world properly pricing and charging for externalities are significant problems: Everyone
attempts to push external costs on others.
To summarize, the model calculates the least cost of meeting demand and the associated competitive
prices. The model contains an electrical component because losses and congestion can only be
calculated with a detailed electrical model. The model contains a detailed description of generator
costs and capabilities because it is necessary to tradeoff costs, congestion and losses across all
potential suppliers to calculate system marginal cost and least cost.
3. The Cost and Price Consequences of Area Interchange Limits
This section reports the estimated impact on system operating costs and prices of area interchange
limits. That amounts to calculating the differences in prices and system costs between the “restricted
trade” regime and the “unrestricted trade” regime. In the restricted trade regime, the relevant demands
and supplies are primarily those within each region. There are some imports and exports but they
cannot exceed the area interchange limits specified in the NERC 2000 case.23 Each region is
minimizing its own costs using its own resources but not fully exploiting further cost reductions
through trade.
In the unrestricted trade regime, the relevant demands and supplies are those across the Eastern
Interconnect and there are no administratively imposed trade restrictions. Trade is, however, restricted
by the physical capabilities of the transmission system. The prices that are calculated in both cases are
those appropriate to a competitive market.
We made cost estimates for summer peak and for a “shoulder” period. The summer peak is NERC’s
estimate of demand when all areas are simultaneously in a heat wave. At such times, essentially all
available generators are on line and there is little room for cost reduction. The shoulder period is
nominally defined as 80% of peak demand. During a shoulder period generation capacity exceeds
22
Calculation and enforcement of congestion charges varies across the regulated jurisdictions in the Eastern Interconnect.
The FERC 715 files specifies the total imports or exports for each area, and in rare instances bilateral transactions are
modeled.
23
8
demand so there is more potential for reducing costs by calling on relatively lower cost generators.
Since demand is reduced in all areas at the same time, and almost all areas have some low cost
generators, there may be limited need for any area to import cheaper power.
The restricted trade case has total operating cost of $10,968,547/hour.24 LMPs were computed at
slightly more than 31,000 buses. The LMPs ranged from a high of $ 987/MWh to a low of $ 16/MWh with less than 6% having values above $ 100/MWh and only eight buses having negative
values.25 The average LMP was $ 57.96/MWh while the standard deviation was $33.71/MWh. The
area interchange constraints are what permit the area average LMPs to vary considerably. They range
from a high of $ 149.82/MWh to a low of $ 6.49/MWh. At the solution the flows on 14 lines were
being enforced as binding constraints; this represents only about 0.05% of the 41,000 lines modeled in
OPF areas.
Next the system was modeled assuming unrestricted trade, i.e., the area interchange contingency
constraints between the U.S. operating areas were relaxed (those with Canadian areas were still
enforced). Thus the entire U.S. portion of the Eastern Interconnect was treated as though it were a
single operating area.26
As expected, relaxing the area contingency constraints at peak demand resulted in a lower operating
cost, with the value dropping by about 3.7% from $11 million/hour to $10.6 million/hour. With the
area constraints relaxed, power was free to flow from the low cost areas to those with higher costs.
The largest absolute changes occurred in the Virginia Power control area, which increased its exports
by over 1,400 MW, while TVA and the Florida Control area saw their imports increase by 1,161 and
1,674 MW respectively.
This resulted in the average area LMPs
converging as well, with the highest value
now just $76.87/MWh and the lowest $
52.46/MWh.27 Compared to the administered
trade model, the average LMP28 remained
about the same, $57.95/MWh. The standard
deviation of the LMPs, however, decreased
substantially
from
$33.71/MWh
to
$8.44/MWh. The price leveling would have
been greater except that increased trade led to
the congestion of 17 lines compared to 14 in
the restricted trade case. Figure 2, which
24
NewIND100d010402.pwb.
Negative LMPs simply indicate that buses on one side of a constraint should reduce generation or increase their load.
Negative LMPs occurred on the PJM system during each of the three summer months in 1999, reaching
$-199.33MWh in July. One area, Consumers Energy in Michigan has an additional 15 LMPs with negative values,
stemming from large reactive power loads.
26 NewSA100d010402.pwb.
27 One nonutility generator, modeled as an area with negligible load, solves with an average LMP of $31.20/MWh.
28Not load weighted. Median LMP cost drops slightly.
25
9
graphs LMPs at about 31,000 locations under both trade regimes, illustrates the degree to which prices
are flattened by free trade.
Area interchange limits at peak demand account for about 4% of operating costs. They also permit
high price pockets throughout the Eastern Interconnect. Recall that the short term operating cost of
area interchange limits is the difference between the least cost of meeting demand while enforcing the
limits and the least cost of meeting demand consistent
Table 6. Cost of Meeting Demand under Two Trade Regimes
with the transmission system’s physical capabilities.
(millions $)
In reality the current balkanized transmission system
Demand Levels
Cost Reductions
Peak
Shoulder
($)
%
Regime
is unlikely to minimize costs within regions.
Restricted
$ 11.0 $
7.6 $
3.4
31%
Consequently this estimate of the cost of area
Unrestricted
$ 10.6 $
7.1 $
3.5
33%
Reductions
interchange limits understates actual costs.
($)
Savings (%)
$
0.4 $
3.6%
0.5
6.6%
Not surprisingly the costs of interchange limits
increase during periods of shoulder demand (Table 6). When peak load is reduced by 20% (to about
404 GW), the cost of meeting demand falls over 30% in both trade regimes. System operating cost
under restricted trade is about $7.6 million per hour and under free trade is about $7.1 million. The
cost of restricting trade is about $500 thousand/hour, about a 25% increase over the costs of
restrictions at peak. Because the cost of restricting trade increases at the same time total costs decrease
the cost of interchange limits as a percent of system operating costs increases from 3.6% to 6.6%.
Power flows generally proceed west to east,
and north to south as the model seeks to
minimize cost differences. The largest net
exporters are TVA (4.4 GW), Duke (3 GW),
and Southern Company (2 GW). Large net
importers are Entergy (5.3 GW), CPLE (4
GW), FRCC (2.9 GW), Commonwealth
Edison (2.6 GW), and ISNE (1.3 GW).
Marginal cost in Entergy falls about
$23/MWh, or 43%, while costs in Ameren
increases by about $4/MWh (Figure 3). Note
that, in this case, load in Ameren is only half
that served in Entergy, indicating the value of
the optimal power transactions.
Figure 3. Area Differences in Marginal Cost After Lifting Trade Restrictions, Shoulder Demand
(Area Loads > 3,000 MW)
10
Higher Cost Areas
5
4.08 3.83 3.79
2.99
2.7
2.67 2.62 2.26
1.26 0.92
0.03
0
-0.08 -0.48
-0.62
-5
-1.19
-2.12 -2.58
-3.85
-4.95
-6.5
-10
-9.11
Lower Cost Areas
-12.35
-15
-20
-25
-22.79
4. Competitive Consequences of Area Interchange Limits: Examples from New England and
Florida
This section shows that the choice of interchange limits can have a major impact on prices by limiting
competition. New England and Florida are used as hypothetical examples. They were chosen in part
because they are in established NERC regions (NPCC and FRCC respectively) that operate under
NERC interchange limits. They differ in their physical ability to import power and in the mix of
generators within their borders, leading to major differences in the impact of interchange
(contingency) limits on competition.
10
In the case of New England, the n-1 contingencies limit the flow of imports to the region off-peak
(Table 7). When imports are capped at the contingency level, 2,526 MW, generation of 16,200 MW is
necessary to meet the ISO demand of 18,299 MW.
Table 7. Key results from New England Cases
Losses total 410 MW. These constraints result in a
Marginal Cost Imports Revenue
ISNE
(MW)
(000)
marginal cost of $36.17/MWh. The minimum LMP
Case Description
Off-peak demand, restricted trade
36.17
2,526
588.8
in New England is $9.17, and the maximum LMP is
Off-peak, restricted, NE offers up 20%
43.52
2,526
707.5
$53.65.
Off-peak demand, optimized trade
27.39
4,016
401.5
Off-peak, with trade, NE offers up 20%
29.69
4,355
423.3
Suppose generator owners, confident that the
contingency constraints would limit imports to about 2500 MW, were to independently increase their
bids by 20% over (marginal) cost.29 The average price would increase to $43.52/MWh, the standard
deviation of prices would increase to $2.26/MWh, the maximum local price increases to $65.04/MWh
and the minimum is $11.08/MWh. Imports remain fixed at 2,526 MW, and generator revenue
increases to $707,477 per hour. Since lines are at their area
interchange limit imports cannot increase, and since demand does not react to price, generators could
continue to increase their bids, consequently raising prices and profits.
Under unrestricted trade and competition, New England (Connecticut, Massachusetts, Maine,
Vermont and New Hampshire) generators produce only 14,706 MWh, a 9 percent decrease.30
Examination of the supply curve shows this level of production has a marginal cost of $27.39 per
MWh, which is the average competitive price.
Eliminating the line constraints reduces the average price (marginal cost) 24% compared to the
corresponding restricted case. Competition and trade reduces price 37%, from $43.52 to $27.39,
compared to the non-competitive case. Differences in the impact of local demand on congestion and
line loss cause local prices to be dispersed around the average. The maximum local price is $39.59, the
minimum is $9.61, and the standard deviation, a measure of spread of local prices about the average, is
$2.02. Generator revenue is $401,492 per hour. The supply curve implies that a price of $27.39 is not
sufficient to induce generation within New England such that demand is met and line losses accounted
for, so the model optimizes by importing 1,516 MW over the planning contingency of 2,526 MWh.
Imports at these levels reduce marginal cost in New England by about 24%.
If generators attempt to raise prices by bidding 20% over (marginal) cost, imports would increase to
4,355 MWh and average price would drop to $29.69.31 Imports are not at their physical limit-more
could actually be imported- because at the margin it would cost more to increase output outside of
New England than would be saved within New England. Absent the contingency constraint, generator
revenues would fall to $423,328 per hour and costs would fall because output declines to 14,391
MWh. Prices would still exceed the competitive standard of $27.39 but not by as much as when the
contingency constraints are applied.
29
Contact Thomas Leckey for details of this case (NewIND80NE20d010402.pwb).
Contact Thomas Leckey for details of this case (NewSA80d010402-a.pwb).
31Contact Thomas Leckey for details of this case (NewSA80NE20d010402-a.pwb).
30
11
Noncompetitive bidding would increase total profits for New England generators above the
competitive level, but some higher cost generators that operated profitably under restricted trade
conditions would be idle. Since the highest bidders are the ones that pay the price in lost profit, they
would have substantial reason to bid true marginal cost. If they “break ranks,” prices will fall. These
pressures will remain until all of the generators threatened with loss of output bid their true (marginal)
costs. In this sense, free trade helps enforce competitive bidding on generators.
Like New England, Florida imports to the level of its contingency limit. Consequently, were
generators to increase bids 20 percent above marginal cost, prices and profits would increase by nearly
the full amount (Table 8). Florida’s consumers get less relief from imports than New England
consumers. Floridians face higher cost generators and have less recourse to cheap imports. Under
unrestricted trade, Florida generators
produce 24,805 MWh,32 resulting in an
Table 8. Key results from Florida Cases
Marginal Cost
Imports
Revenue
average competitive price of about $44.97
(MW)
(000)
FRCC
Case Description
MWh with a standard deviation of $7.32.
2,268
1,308
Off-peak demand, restricted trade
47.52
2,268
1,572
Off-peak, restricted, FL offers up 20%
57.09
Imports total 5,186 MWh.
Off-peak demand, optimized trade
Off-peak, with trade, FL offers up 20%
44.97
54.31
5,186
5,188
1,088
1,312
Were Florida generators to raise their offers
by 20%, prices would increase to $54.31 MWh.33 Prices essentially increase as much as offers, but
Florida generation is almost unchanged because the transmission system does not permit additional
imports. Given that additional imports are not available and demand does not respond to price
increases, generators can raise their prices and increase profits at will.
Florida’s electricity market is vulnerable to opportunistic pricing in the short run even with
unrestricted trade. The summer limits exacerbate a difficult problem by constraining imports even
during non-peak times. In the long term electrical connections could be strengthened to allow more
imports. Perhaps lower cost generators could be built. But, in the short run competitive trade is not
enough to force competitive pricing.
5. Implications of Load Pockets: New York City
Even if area interchange limits were eased, some local areas would continue to face persistently high
costs by virtue of their location in the existing transmission grid. New York City, with just under 11
GW of load at peak, constitutes about 40% of peak load in the New York ISO. In 2000 the city was
served by four large fossil plants sited within the city totaling about 5 GW generating capacity, and by
three utility transmission interfaces which are capable of supplying the remainder of peak demand.
32
33
Contact Thomas Leckey for details of this case (NewSA80d010402-a.pwb).
Contact Thomas Leckey for details of this case (NewSA80FL20d010402-a.pwb).
12
When trade restrictions are in place, LMPs in New
York City run about 9% higher compared to an
open trade regime (Figure 5).34 The graph depicts
64 load bearing buses in the city under restricted
and free trade, where about 9 GW of demand is
present. When trade is introduced, the average
cost at these 64 buses declines from $27.64 MWh
to $25.28 MWh. Loadings on the thirteen main
lines to the city approach capacity, a significant
portion of which is dedicated to reactive power.
Despite the cost reduction that comes along with
increased trade, costs within the city remain
significantly higher than in the NYISO as a whole
(Figure 6). By means of a load normalized curve,
the city’s 9 GW of load is compared to the nearly
23 GW of load in the ISO.35 Costs at the city load
buses exceed those in the ISO for the first 80% of
load, after which the high cost buses elsewhere in
the State outstrip the costs in the city. Because
access to the city through the existing interfaces is
constrained, the area would remain vulnerable to
market power, even with broad interregional trade.
This situation is similar to the physical constraints
observed in Florida.
7. Conclusion
Area interchange limits based on n-1 contingency studies have been an effective weapon for 40 years
in limiting area blackouts in the Eastern Interconnection. However, area interchange limits are blunt
tools, insensitive to the actual risks and opportunities facing the system in real time and ignorant of
economic consequences.
This paper has demonstrated that the hidden costs of these limits can also be significant. Operating
costs could be reduced on the order of 4-7% if the network were used to its physical potential. Subtler
is the finding that area interchange constraints can shield noncompetitive pricing and inefficient
operations. In the case of New England this shield is decisive. Even were New England’s market
perfectly competitive, the average price under area interchange limits would be 30% above those
prevailing under free trade. Florida’s consumers are disadvantaged by area interchange limits but
removing them is not enough to compensate for a weak inter-regional transmission system and high
cost generation. Florida generators are well protected either from pressures to economize or to price at
34
NewSA80d010402-a.pwb. About 80% of peak load.
The x-axis arranges the approximately 700 load-bearing buses in the ISO according to their share of the 23 GW.
Similarly, each of the 64 city load buses constitute a share of the 9 GW load, so that both curves extend the full length of
the x-axis. The 64 NYC buses are also represented in the ISO curve.
35
13
marginal cost. New York City’s consumers face a constrained transmission system, resulting in prices
higher than those obtained in the remainder of the ISO.
The tools for detecting, quantifying, and communicating risk are far advanced from what they were 40
years ago. And, the cost and competitive consequences of interchange limits are significant. It is
probable that significant savings and efficiencies can be realized by re-examining reliability tools in
the light of changed technology and market realities.
8. Acknowledgement and References
Research for the generator cost database was conducted by Ms. Marilyn Walker of the Department of
Justice. A table matching NERC Names and reference numbers to those used by the EIA and U.S.
Environmental protection agency are available from the authors on request.
9. Biographical Information
Douglas Hale (202.287.1723) is a senior economist at the Energy Information Administration (EIA) of
the U.S. Department of Energy; Thomas Overbye is a professor of electrical engineering, University
of Illinois; and Thomas Leckey (202.586.9413) is an industry specialist at EIA.
14